Proportionally Fair Clustering

The data points in machine learning are often real human beings. There is legitimate concern that traditional machine learning algorithms that are blind to this fact may inadvertently exacerbate problems of bias and injustice in society [25]. Motivated by concerns ranging from the granting of bail in the legal system to the quality of recommender systems, researchers have devoted considerable effort to developing fair algorithms for the canonical supervised learning tasks of classification and regression [13, 28, 20, 27, 34, 11, 30, 35, 26, 18, 21]. We extend this work to a canonical problem in unsupervised learning: centroid clustering. In centroid clustering, we want to partition data into k clusters by choosing k "centers" and then matching points to one of the centers.